Customer Service Telegram Bot တစ်ခု တည်ဆောက်မယ်။

• Next.js
• Telegram Bot
• Open Router AI
• In Memory Storage for session

---

Prompt

📌 AI Agent Prompt: Build FTTH Provider Telegram Bot

Role: You are an expert Next.js + Telegram Bot + AI integration developer. Goal: Develop a Telegram chatbot in Next.js for an FTTH provider in Myanmar, with multi-turn conversations to create and query support tickets.

---

🟢 Requirements

1. Framework & Tools

   • Use Next.js (latest stable).
   • Create API routes in /pages/api/telegram/webhook.ts for handling Telegram updates.
   • Use axios for external API calls.
   • Use ngrok for local development webhook.

2. Telegram Bot

   • Receive messages via Telegram webhook.
   • Send replies via sendMessage.

3. Conversation Flow (Myanmar language)

   • Multi-turn chat, maintained per user session (chatId).

   • Intent choices:

     • create_ticket
     • query_ticket

   • Conversation steps:

     • awaiting_intent → Ask: “Ticket အသစ်ဖွင့်ချင်ပါသလား၊ သို့မဟုတ် ရှိပြီးသား Ticket စစ်ချင်ပါသလား?”
     • awaiting_customerId → Ask for Customer ID.
     • awaiting_issue (only for create) → Ask for issue description.
     • done → Call CRM API and return result.

4. AI Layer (OpenRouter / DeepSeek)

   • Use OpenRouter API (https://openrouter.ai/api/v1/chat/completions).

   • Model: deepseek-chat.

   • Prompt AI to map Myanmar free-text → JSON:

     { "intent": "create_ticket" }
     { "intent": "query_ticket" }
     { "intent": "create_ticket", "customerId": "...", "issue": "..." }
     

   • Example system prompt:

     You are an AI assistant for an FTTH provider in Myanmar.
     Detect whether the user wants to "create_ticket" or "query_ticket".
     Extract fields: customerId, issue (if provided).
     Always reply in valid JSON only.
     

5. CRM API Integration

   • For creating a ticket: POST http://crm.local/api/tickets { customerId, issue }
   • For querying tickets: GET http://crm.local/api/tickets?customerId=xxx

   • Assume response:

     { "ticketNumber": "TCK-1234" }
     

     or

     [{ "ticketNumber": "TCK-1234", "status": "open" }]
     

6. Session Management

   • Use in-memory Map (sessions[chatId]) to store:

     {
       step: "awaiting_intent" | "awaiting_customerId" | "awaiting_issue" | "done",
       intent?: string,
       customerId?: string,
       issue?: string
     }
     

   • Reset session after completion.

7. Local Development

   • Run: ngrok http 3000

   • Register webhook:

     curl -F "url=https://<ngrok-url>/api/telegram/webhook" \
          https://api.telegram.org/bot<TELEGRAM_BOT_TOKEN>/setWebhook
     

---

🟡 Deliverables

• app/api/v1/telegram-bot/route.ts (main bot logic).
• lib/session.ts (optional, session store).

• Example .env file with:

  TELEGRAM_BOT_TOKEN=xxxx
  OPENROUTER_API_KEY=xxxx
  CRM_API_BASE=http://crm.local/api
  

• Instructions for running locally with ngrok.

---

🔵 Expected Behavior

• User: “ကျွန်တော့် Wi-Fi မအလုပ်လုပ်ဘူး”

• Bot: “မိမိ Customer ID ပေးပါ။”

• User: “testee@myanmarlink.net”

• Bot: “Ticket ဖွင့်ပြီးပါပြီ။ နံပါတ်: TCK-1234”

• User: “ကျွန်တော့် Ticket အခြေအနေ စစ်ချင်တယ်”

• Bot: “မိမိ Customer ID ပေးပါ။”

• User: “testee@myanmarlink.net”

• Bot: “သင့်မှာ Ticket 2 ခုရှိပါတယ် – TCK-1234: ဆက်လုပ်ဆဲ, TCK-1201: ပြီးဆုံးပြီး။”

---

✅ Build this project step by step. Generate production-ready, clean, and well-documented code.